Configuration Reference

Configuration Reference

All non-sensitive, runtime configurations are managed via a YAML-based config loader. The primary config file is apps/backend/config.default.yaml.

To override these defaults for specific environments, you can create:

• apps/backend/config.dev.yaml (automatically loaded in development)
• apps/backend/config.prod.yaml (automatically loaded in production)

Important

Secrets and Keys: Sensitive credentials (database URIs, API keys, passwords, webhook tokens) must never be committed to YAML files. Keep them in your .env or inject them at runtime using environment variables.

---

Configuration Parameter Sections

Below is a detailed breakdown of the main config sections:

1. Server Configuration (server)

• port: Port the Express server listens on (defaults to 6767).
• trustProxyHops: Number of reverse proxy hops to trust (critical for tracking correct user IPs via Nginx).
• bodyLimit: Maximum size of incoming HTTP bodies (e.g. 10mb).
• shutdownTimeoutMs: Graceful shutdown timeout (defaults to 15000ms).

2. CORS Settings (cors)

• allowedOrigins: List of origins allowed to perform cross-origin requests.
• allowVercelPreviews: Boolean toggle to trust dynamic preview domains.
• allowLocalhostInDev: Automatically permits localhost origins when running in development mode.

3. JWT and Authentication (auth)

• jwt.expiresIn: Duration before the authentication token expires (default 7d).
• jwt.issuer: The token issuer string (default csfaq).
• jwt.audience: The target API audience (default csfaq-api).
• password.bcryptRounds: BCrypt work factor (default 10 rounds).

4. Rate Limiting (rateLimiting)

Governs IP-based request throttles using express-rate-limit:

• global: Global traffic limits.
• auth: Stricter thresholds for login (max 5 in 15m) and registration (max 3 in 60m).
• search: Limits search requests (max 30 per minute).
• askAi: Daily AI quotas (max 5 for anonymous searches, 50 for logged-in students).

5. Search Engine (search)

• hybrid.vectorWeight: Relative weight of vector rankings in RRF merging (default 0.6).
• hybrid.keywordWeight: Relative weight of keyword text rankings in RRF merging (default 0.4).
• hybrid.rrfK: The constant $k$ used to damp Reciprocal Rank Fusion ranks (default 60).
• hybrid.minScore: Minimum score similarity threshold below which results are discarded (default 0.3).
• embedding.model: Hugging Face model key used for local embedding generation (default mixedbread-ai/mxbai-embed-large-v1).
• embedding.dimensions: The number of vector dimensions (default 1024).

6. FAQ and Freshness Tiers (faq)

• freshness.tiers: Review intervals before content is flagged for verification:
  • Evergreen: 90 days.
  • Seasonal: 15 days.
  • Volatile: 4 days.
• freshness.peerVoteThreshold: Number of helpful votes needed from experts to auto-verify freshness.
• duplicateDetection.similarityThreshold: Math threshold above which a submission is auto-flagged as a duplicate (default 0.85).

7. Support Request Settings (support)

• goldenTicket.spCostPerEscalation: Points deducted to make a support request high-priority (default 10).
• goldenTicket.defaultCooldownHours: Delay before a student can request another Golden Ticket (default 48h).
• troubleshoot.maxSteps: Steps defined for the student troubleshooting guides (default 4).

8. AI Pipelines (ai)

• providers.priority: Order of provider lookup when resolving fallback services (default: anthropic, openai, xai, minimax, gemini).
• pipelines.autoAnswer.approveThreshold: Confidence score above which the AI will automatically answer a community question without human moderation (default 0.85).
• pipelines.autoAnswer.queueThreshold: Score above which the AI answer is queued for review (default 0.60).

9. Reputation and Tiers (reputation)

• points: Points awarded to students:
  • Creating a community post: +5 points.
  • Creating a comment: +2 points.
  • Having their answer accepted: +15 points.
  • Receiving an upvote: +1 point.
  • Contributing a verified FAQ: +10 points.
• tiers: Reputation milestones:
  • Newcomer: 0 points.
  • Contributor: 50 points.
  • Expert: 200 points.
  • Mentor: 500 points.
  • Knowledge Master: 1000 points.

10. Data Retention (retention)

Governs clean-up intervals for logs and records:

• searchLogs.ttlDays: 90 days.
• notifications.ttlDays: 90 days.
• freshReviewLogs.ttlDays: 180 days.
• moderationLogs.ttlDays: 365 days.